lly form a direct salt-bridge to Asp235 has become higher in energy than an orientation that maximizes their interactions with the new water network. This is a point to which we return. Prospective Docking against the Open Cavity To investigate the ability of docking to predict new ligands of the Gateless cavity, 534,000 molecules with molecular weight between 30 and 250 Da were docked against its structure. Using a variation of an receiver operator curve that plots log10 of the percent of decoys found on the x-axis, which acts to up-weight early ligand enrichment, and that corrects for the enrichment seen at random, logAUC32, high enrichment of known ligands was observed, with the three sub-millimolar ligands in the top 1.4% of the database. Whereas the average molecular weight of the six known ligands for Gateless is 112 Da, larger compounds dominated the top of the docking list, with an average molecular weight of 210 Da for the first 500 compounds. Intriguingly, a substantial number of highly-ranked compounds were uncharged . We were skeptical of this result, given the very modest affinities of phenol and 3-fluorocatechol for this site, and so turned to a Solvent-Excluded Volume method of accounting for ligand desolvation in docking. This method calculates the amount of solvent dielectric excluded by the volume of lowdielectric protein for any given configuration of a ligand in a binding site, using this value to calculate the ligand desolvation through a version of the Born equation. In this treatment a ligand 3 Affects of Water on Ligand Recognition and Docking is more or less desolvated depending on the overall volume of protein that surrounds it, and even a fully buried ligand retains an interaction with the bulk, remaining partially solvated, as is physically correct. This SEV method provides a better model than either considering a ligand fully desolvated on docking to a site which over-penalizes itor not desolvating it at all, which underpenalizes it, and does so in a physics-based manner consistent with the rest of the DOCK3.6 scoring function. In retrospective studies, this SEV method had improved enrichments for sites where the 
solvent interface plays an important role and also appeared to do so in a prospective screen; this study represents the first test of the method, in a model system that enables detailed analysis of the results. TMS biological activity Redocking the ZINC library with the SEV method, the fraction of neutral compounds in the top 5000 molecules of the docking list dropped to 3% and enrichment for active compounds improved slightly. This fit with our expectation that this site, though opened to solvent, would still be dominated by cationic ligand recognition. To actually test this, three neutral molecules, compounds 7, 8 and 9, that were prioritized by the older full desolvation method and de-prioritized by the SEV method were tested for binding; none showed measurable affinity for the Gateless cavity at up to 1 mM concentration. The Gateless cavity offered the first chance to test the new SEV solvation 22634634 method prospectively. A further fifteen molecules, in addition to the three neutral molecules from the previous hit list, were selected from among the top 500 compounds of the SEV solvation-based hit list, or top 0.1% of the database screened, and tested for affinity. In addition to 9521749 the docking score that ranked them among the top 500 molecules, these compounds were selected for favorable interactions with key binding site resid
